Allow for more than ten bits in subpixel precision measuring; modern llvmpipe seems...

[movit] / blue.frag

vec4 FUNCNAME(vec2 tc) {
        return vec4(0.0, 0.0, 1.0, 1.0);
}